Research map / Video / Video generation
RFDM: Residual Flow Diffusion Models for Video Editing
Mohammadreza Salehi, Mehdi Noroozi, Luca Morreale, Ruchika Chavhan, Malcolm Chadwick, Alberto Gil Couto Pimentel Ramos, Abhinav Mehrotra
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| predict residual flow instead of full frame (ablation) | RFDM1.5 · Senorita val, global style transfer | error accumulation: 0.09 -> 0.07 | Improved |
| predict residual flow instead of full frame (ablation) | RFDM · DAVIS tracking (10 videos) | J&F: 29.1 -> 43.6 | Improved |
| add clean previous prediction as extra input (ablation) | RFDM1.5 · Senorita val, global style transfer | TempCon: 0.027 -> 0.009 | Improved |
| increase autoregressive frames in loss from 0 to 5 (ablation) | RFDM1.5 · Senorita val, global style transfer | TempCon: 0.068 -> 0.007 | Improved |
| use diffusion forcing instead of teacher forcing (ablation) | RFDM1.5 · Senorita val, global style transfer | ViDreamSim: 0.38 -> 0.35 | Improved |
| use diffusion forcing instead of teacher forcing (ablation) | RFDM1.5 · Senorita val, global style transfer | error accumulation: 0.06 -> 0.07 | Hurt |
| update previous-frame key frame every 3 frames instead of every frame (ablation) | RFDM1.5 · Senorita val, global style transfer | error accumulation: 0.16 -> 0.07 | Improved |
| never update key frame (delta=0) (ablation) | RFDM1.5 · Senorita val, global style transfer | TempCon: 0.009 -> 0.018 | Hurt |
| unroll gradients through autoregressive stages (ablation) | RFDM1.5 · Senorita val, global style transfer | TempCon: 0.011 -> 0.009 | Improved |
| scale backbone from SD1.5 to SD3.5 | RFDM · Senorita test set | MLLM-Judge: 6.60 -> 7.37 | Improved |
| RFDM causal autoregressive editing vs per-frame I2I baselines | RFDM3.5 · Senorita test set | ViDreamSim: 0.63 -> 0.20 | Improved |
| RFDM vs VidToMe on temporal consistency | RFDM3.5 · Senorita test set | TempCon: 0.007 -> 0.009 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- Fairy — Fairy: Fast Parallelized Instruction-Guided Video-to-Video Synthesis (CVPR 2024) · better on 4 of 4 shared results
- RAVE — RAVE: Randomized Noise Shuffling for Fast and Consistent Video Editing with Diffusion Models (CVPR 2024) · better on 4 of 4 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- FADE: Frequency-Aware Diffusion Model Factorization for Video Editing (CVPR 2025 · similarity 0.91)
- COVE: Unleashing the Diffusion Feature Correspondence for Consistent Video Editing (NeurIPS 2024 · similarity 0.89)
- MeDM: Mediating Image Diffusion Models for Video-to-Video Translation with Temporal Correspondence Guidance (AAAI 2024 · similarity 0.89)
- TokenFlow: Consistent Diffusion Features for Consistent Video Editing (ICLR 2024 · similarity 0.89)
- MaskINT: Video Editing via Interpolative Non-autoregressive Masked Transformers (CVPR 2024 · similarity 0.88)
- VideoGuide: Improving Video Diffusion Models without Training Through a Teacher's Guide (CVPR 2025 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
